Research And Development Of A Portable Heart Sound Detection System

Cardiovascular disease is a common modern disease with sudden onset and high fatality rate,and clinical symptoms of ischemia and hypoxia can easily lead to irreversible damage to multiple organs in the human body.Therefore,timely detection and early warning of cardiovascular disease are extremely important.Heart sounds carry rich information reflecting the state of cardiovascular health.In clinical practice,manual auscultation of heart sounds is often used as a primary screening method for cardiovascular diseases.However,it relies on the clinical experience of doctors and cannot be accurately recorded and archived.With the development of computer-aided diagnosis technology,electronic auscultation of heart sounds has become possible.Many scholars have studied intelligent heart sound detection methods with the goal of high recognition accuracy of abnormal heart sounds.However,the high accuracy of the algorithm often comes at the expense of the high complexity and parameter of the model,which has high demands on the performance of the device and poor detection portability.A portable heart sound detection system based on the mechanism of heart sound generation is developed in this study.Firstly,the function modeling of the detection system is carried out based on Unified Modeling Language to clarify the development process.Then,by analyzing the characteristics of the heart sound signal and its acquisition principle,a hardware acquisition device is designed,and the main components of heart sounds are obtained.In addition,by using the heart sound signals collected autonomously and from the Physio Net database,a lightweight heart sound detection model based on a one-dimensional convolutional neural network is constructed.By building a deep model,and studying lightweight optimization methods such as pruning and knowledge distillation,the parameters of the deep model are compressed,and the rapid detection of abnormal heart sounds is realized.Finally,the lightweight model is initially deployed in mobile phones.An experiment is designed to test the performance of the portable heart sound detection system.According to such indicators as heart sound waveform and heart rate,the effectiveness of the acquisition device was tested.And the effectiveness of the lightweight detection model is tested in terms of accuracy,memory,prediction time,and so on.According to the results,the portable heart sound detection system designed in this paper can effectively collect the main components of heart sounds,and the detection accuracy of heart sounds can reach97.69%.The size of the designed detection model is 392 KB,the detection speed is about 0.0014 seconds,and the floating-point operations are 0.04 GMac.To sum up,the portable heart sound detection system proposed in this paper has the advantages of light weight,high detection accuracy,and fast prediction speed.The designed lightweight model can be deployed in the signal acquisition device in the next step to realize the integration of acquisition and detection.It has important practical significance for intelligent detection of heart sounds in family health care and medically underdeveloped areas.